The interviewer reads the page above and asks: “It's 3 a.m. and this just woke you up. What do you do?”
Both answers were written for this page. Neither comes from a real candidate or interview.
B. You just ran the Pager Test.
Here's what you picked up on, probably without listing it.
A is the one that reads better on paper.
It's organized, complete and correct. Here's what B shows that A doesn't, and what happens when you ask A one more question.
The answer: B.
Here's what gives it away, and what happens when you ask A one more question.
What B shows
- Does the math first. “About forty minutes” sets the pace for everything else.
- Names the usual suspect, and how it works. A replication slot nobody is reading makes the primary keep every WAL file. You don't learn that from a course; you learn it when it fills a disk.
- Fixes the cause, not the symptom. Fix the consumer and the disk drains on its own. More space only buys time.
- Knows the trap in the obvious fix. EBS makes you wait six hours between changes to a volume, so you grow it once, and generously. People learn that once, usually at 3 a.m.
- A rule with a scar behind it. “I don't delete files I can't name.”
- Talks to people before they ask. The update goes out before the first “any update?”
Where A runs out
Interviewer You scaled up storage. What was filling the disk?
A I'd look into that during the root-cause analysis, using our monitoring and logs.
Interviewer It's still climbing a percent every ten minutes. What do you look at first?
A Disk usage metrics, and the database logs for errors.
Interviewer The logs are clean. Then what?
A I'd probably escalate to the database team.
That's the bottom, three questions in. It's a fine answer from a mid-level engineer. It isn't one from the person who is the database team at 3 a.m.
Why this is hard to spot
Answer A isn't wrong. Every step in it is correct, and it's also roughly what a chatbot writes when you ask it the question. On paper, in a take-home or in an AI-polished CV, A and B can look the same. The difference only shows when someone who has been paged keeps asking “and then what?”
That's the Pager Test: a 45-minute interview where the candidate walks through a system they've actually run, and I keep asking until they reach the bottom of what they've done, then write down where that was.
Run it on your real candidates
$2,500
Pressure-test your top pick before the offer. Calibration call, the interview, a written assessment by the next business day, and a debrief.
Book for one →$4,500
Torn between two? Both interviewed at the depth your role needs and assessed against the same bar, so you're comparing evidence, not impressions.
Book for two →